Estimating the spatial

pattern and extent of

hemlock mortality after

HWA infestation in the

Linville River Gorge via

aerial imagery

Sheryl Strauch1, Tuula Kantola2,

Maria Tchakerian1, Robert

Coulson1 & Douglas Streett3

1Knowledge Engineering Laboratory, Dept. of

Entomology, Texas A&M University, College

Station, TX

2Dept. of Forest Sciences, University of

Helsinki, Finland

3USDA Forest Service, SRS, Pineville, LA.

SFIWC 2013

What’s bugging us? Invasive species are often a significant threat

to US Forests

Severe outbreaks cause major changes in landscape structure, function, and composition

Hemlock woolly adelgid’s (HWA) rapid spread and infestation throughout eastern forests

Goal: to understand effects of this invasive species and obtain more data on the impact of HWA because so little known

Data quantifying damages does not exist

Field data is lacking

Gaps created in forest following tree mortality

We need more information about HWA effects!

Study Objectives

1) Estimate number and locations of dead trees in study area

2) Estimate the area covered by dead trees

3) Study the spatial pattern of dead trees

https://gustavus.edu/arboretum/treelist.php

Eastern Hemlock

Long lived and extremely shade tolerant

Foundation species

Habitat for a diverse of number of

species

No resistance to HWA observed

http://www.flickr.com/photos/ozark_pix/243308376/

HWA An invasive species causing high

mortality in Eastern Hemlock

http://www.nps.gov/grsm/naturescience/hemlock-woolly-adelgid.htm

Study Area

The Linville River Gorge, Pisgah National Forest, North Carolina

~26 km² area

Upper half of the Lower Linville River watershed

Known presence of Eastern hemlock and HWA infestation

0.4 km2 of shadows were excluded from analysis

Data 1m resolution CIR Imagery-2012

CIR is highly useful in forest assessment

Actively growing plants exhibit high NIR reflectance

Differentiating species

Distinct colors for identifying live vs. dead

15cm resolution reference RGB aerial photo imagery-2010

RGB CIR

Methods: Dead Tree Identification All dead trees were located and digitized with points

A random 10% subsample was taken and used for crown area

estimation via polygon creation

Total area covered by dead trees was estimated using

distribution of crown polygon areas by entire population

Assumed constant distribution throughout the population

Dead tree crowns

Results: Dead Tree Identification 6465 dead trees identified and located (assumed dead trees to be

hemlocks)

647 tree crown areas digitized

Crown area distribution (assumed population has same distribution

throughout)

Average dead tree crown size of 24 m2

Dead trees cover 0.6% of the study area

0

20

40

60

80

100

120

140

5 10 15 20 25 30 35 40 45 50 55 60

Fre

qu

ency

Crown Area Size Class (m²)

Methods: Spatial Pattern Analysis

Grid of 1 sq. km cells 36 subareas

Ripley’s K function Tests for deviation from (CSR), complete spatial

randomness

Describes spatial pattern over different spatial

scales

Null hypothesis: K function of the point pattern

does not vary significantly from the homogenous

Poisson process

Linearized K function (L function) Easier to interpret

Variance of all values of t in L function is

almost constant

Results: Spatial Pattern Clustering at nearly all scales

L(t) = 0, pattern is random

L(t) > 0, pattern is clustered

L(t) < 0, pattern is regular

Number of dead trees per subarea ranging 10 to 839

Sample size: 768

Results: Spatial Pattern

Sample size: 494

Conclusions Dead trees covered 0.6% of the study area

The spatial pattern of dead trees was clustered

This may provide useful evidence in the future spread and mortality pattern by HWA

Distribution skewed to the right

We think because of understory trees not being visible, thus not as many identified

No seedling or young stands present

Results help estimate future forest gaps

Estimating impacts of invasive species

Ground truth data needed to assess accuracy of visual identification of hemlocks

Questions?

http://mdmpix.com/2009/2009-10-21_003/